hoernschen/dendrite
1
0
Fork 0
mirror of https://github.com/hoernschen/dendrite.git synced 2024-12-26 15:08:28 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Move the SQL into separate files for each table. (#7)

Browse source 
* Move partition offsets sql to a separate file

* Move event types sql to a separate file

* Move event state key sql to a separate file

* Move room sql to a separate file

* Move events sql to a separate file

* Move event json sql to a separate file

* Move state snapshot sql to a separate file

* Move state block sql to a separate file
This commit is contained in:
Mark Haines 2017-02-15 14:43:19 +00:00 committed by GitHub
parent 39264cbf4b
commit 450e0f1570
9 changed files with 846 additions and 754 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

93
src/github.com/matrix-org/dendrite/roomserver/storage/event_json_table.go Normal file
View file

@ -0,0 +1,93 @@
package storage
import (
"database/sql"
"github.com/lib/pq"
"github.com/matrix-org/dendrite/roomserver/types"
)
const eventJSONSchema = `
-- Stores the JSON for each event. This kept separate from the main events
-- table to keep the rows in the main events table small.
CREATE TABLE IF NOT EXISTS event_json (
-- Local numeric ID for the event.
event_nid BIGINT NOT NULL PRIMARY KEY,
-- The JSON for the event.
-- Stored as TEXT because this should be valid UTF-8.
-- Not stored as a JSONB because we always just pull the entire event
-- so there is no point in postgres parsing it.
-- Not stored as JSON because we already validate the JSON in the server
-- so there is no point in postgres validating it.
-- TODO: Should we be compressing the events with Snappy or DEFLATE?
event_json TEXT NOT NULL
);
`
const insertEventJSONSQL = "" +
"INSERT INTO event_json (event_nid, event_json) VALUES ($1, $2)" +
" ON CONFLICT DO NOTHING"
// Bulk event JSON lookup by numeric event ID.
// Sort by the numeric event ID.
// This means that we can use binary search to lookup by numeric event ID.
const bulkSelectEventJSONSQL = "" +
"SELECT event_nid, event_json FROM event_json" +
" WHERE event_nid = ANY($1)" +
" ORDER BY event_nid ASC"
type eventJSONStatements struct {
insertEventJSONStmt *sql.Stmt
bulkSelectEventJSONStmt *sql.Stmt
}
func (s *eventJSONStatements) prepare(db *sql.DB) (err error) {
_, err = db.Exec(eventJSONSchema)
if err != nil {
return
}
if s.insertEventJSONStmt, err = db.Prepare(insertEventJSONSQL); err != nil {
return
}
if s.bulkSelectEventJSONStmt, err = db.Prepare(bulkSelectEventJSONSQL); err != nil {
return
}
return
}
func (s *eventJSONStatements) insertEventJSON(eventNID types.EventNID, eventJSON []byte) error {
_, err := s.insertEventJSONStmt.Exec(int64(eventNID), eventJSON)
return err
}
type eventJSONPair struct {
EventNID types.EventNID
EventJSON []byte
}
func (s *eventJSONStatements) bulkSelectEventJSON(eventNIDs []types.EventNID) ([]eventJSONPair, error) {
nids := make([]int64, len(eventNIDs))
for i := range eventNIDs {
nids[i] = int64(eventNIDs[i])
}
rows, err := s.bulkSelectEventJSONStmt.Query(pq.Int64Array(nids))
if err != nil {
return nil, err
}
defer rows.Close()
// We know that we will only get as many results as event NIDs
// because of the unique constraint on event NIDs.
// So we can allocate an array of the correct size now.
// We might get fewer results than NIDs so we adjust the length of the slice before returning it.
results := make([]eventJSONPair, len(eventNIDs))
i := 0
for ; rows.Next(); i++ {
result := &results[i]
var eventNID int64
if err := rows.Scan(&eventNID, &result.EventJSON); err != nil {
return nil, err
}
result.EventNID = types.EventNID(eventNID)
}
return results[:i], nil
}

98
src/github.com/matrix-org/dendrite/roomserver/storage/event_state_keys_table.go Normal file
View file

@ -0,0 +1,98 @@
package storage
import (
"database/sql"
"github.com/lib/pq"
"github.com/matrix-org/dendrite/roomserver/types"
)
const eventStateKeysSchema = `
-- Numeric versions of the event "state_key"s. State keys tend to be reused so
-- assigning each string a numeric ID should reduce the amount of data that
-- needs to be stored and fetched from the database.
-- It also means that many operations can work with int64 arrays rather than
-- string arrays which may help reduce GC pressure.
-- Well known state keys are pre-assigned numeric IDs:
-- 1 -> "" (the empty string)
-- Other state keys are automatically assigned numeric IDs starting from 2**16.
-- This leaves room to add more pre-assigned numeric IDs and clearly separates
-- the automatically assigned IDs from the pre-assigned IDs.
CREATE SEQUENCE IF NOT EXISTS event_state_key_nid_seq START 65536;
CREATE TABLE IF NOT EXISTS event_state_keys (
-- Local numeric ID for the state key.
event_state_key_nid BIGINT PRIMARY KEY DEFAULT nextval('event_state_key_nid_seq'),
event_state_key TEXT NOT NULL CONSTRAINT event_state_key_unique UNIQUE
);
INSERT INTO event_state_keys (event_state_key_nid, event_state_key) VALUES
(1, '') ON CONFLICT DO NOTHING;
`
// Same as insertEventTypeNIDSQL
const insertEventStateKeyNIDSQL = "" +
"INSERT INTO event_state_keys (event_state_key) VALUES ($1)" +
" ON CONFLICT ON CONSTRAINT event_state_key_unique" +
" DO UPDATE SET event_state_key = $1" +
" RETURNING (event_state_key_nid)"
const selectEventStateKeyNIDSQL = "" +
"SELECT event_state_key_nid FROM event_state_keys WHERE event_state_key = $1"
// Bulk lookup from string state key to numeric ID for that state key.
// Takes an array of strings as the query parameter.
const bulkSelectEventStateKeyNIDSQL = "" +
"SELECT event_state_key, event_state_key_nid FROM event_state_keys" +
" WHERE event_state_key = ANY($1)"
type eventStateKeyStatements struct {
insertEventStateKeyNIDStmt *sql.Stmt
selectEventStateKeyNIDStmt *sql.Stmt
bulkSelectEventStateKeyNIDStmt *sql.Stmt
}
func (s *eventStateKeyStatements) prepare(db *sql.DB) (err error) {
_, err = db.Exec(eventStateKeysSchema)
if err != nil {
return
}
if s.insertEventStateKeyNIDStmt, err = db.Prepare(insertEventStateKeyNIDSQL); err != nil {
return
}
if s.selectEventStateKeyNIDStmt, err = db.Prepare(selectEventStateKeyNIDSQL); err != nil {
return
}
if s.bulkSelectEventStateKeyNIDStmt, err = db.Prepare(bulkSelectEventStateKeyNIDSQL); err != nil {
return
}
return
}
func (s *eventStateKeyStatements) insertEventStateKeyNID(eventStateKey string) (types.EventStateKeyNID, error) {
var eventStateKeyNID int64
err := s.insertEventStateKeyNIDStmt.QueryRow(eventStateKey).Scan(&eventStateKeyNID)
return types.EventStateKeyNID(eventStateKeyNID), err
}
func (s *eventStateKeyStatements) selectEventStateKeyNID(eventStateKey string) (types.EventStateKeyNID, error) {
var eventStateKeyNID int64
err := s.selectEventStateKeyNIDStmt.QueryRow(eventStateKey).Scan(&eventStateKeyNID)
return types.EventStateKeyNID(eventStateKeyNID), err
}
func (s *eventStateKeyStatements) bulkSelectEventStateKeyNID(eventStateKeys []string) (map[string]types.EventStateKeyNID, error) {
rows, err := s.bulkSelectEventStateKeyNIDStmt.Query(pq.StringArray(eventStateKeys))
if err != nil {
return nil, err
}
defer rows.Close()
result := make(map[string]types.EventStateKeyNID, len(eventStateKeys))
for rows.Next() {
var stateKey string
var stateKeyNID int64
if err := rows.Scan(&stateKey, &stateKeyNID); err != nil {
return nil, err
}
result[stateKey] = types.EventStateKeyNID(stateKeyNID)
}
return result, nil
}

94
src/github.com/matrix-org/dendrite/roomserver/storage/event_types_table.go Normal file
View file

@ -0,0 +1,94 @@
package storage
import (
"database/sql"
"github.com/matrix-org/dendrite/roomserver/types"
)
const eventTypesSchema = `
-- Numeric versions of the event "type"s. Event types tend to be taken from a
-- small common pool. Assigning each a numeric ID should reduce the amount of
-- data that needs to be stored and fetched from the database.
-- It also means that many operations can work with int64 arrays rather than
-- string arrays which may help reduce GC pressure.
-- Well known event types are pre-assigned numeric IDs:
-- 1 -> m.room.create
-- 2 -> m.room.power_levels
-- 3 -> m.room.join_rules
-- 4 -> m.room.third_party_invite
-- 5 -> m.room.member
-- 6 -> m.room.redaction
-- 7 -> m.room.history_visibility
-- Picking well-known numeric IDs for the events types that require special
-- attention during state conflict resolution means that we write that code
-- using numeric constants.
-- It also means that the numeric IDs for common event types should be
-- consistent between different instances which might make ad-hoc debugging
-- easier.
-- Other event types are automatically assigned numeric IDs starting from 2**16.
-- This leaves room to add more pre-assigned numeric IDs and clearly separates
-- the automatically assigned IDs from the pre-assigned IDs.
CREATE SEQUENCE IF NOT EXISTS event_type_nid_seq START 65536;
CREATE TABLE IF NOT EXISTS event_types (
-- Local numeric ID for the event type.
event_type_nid BIGINT PRIMARY KEY DEFAULT nextval('event_type_nid_seq'),
-- The string event_type.
event_type TEXT NOT NULL CONSTRAINT event_type_unique UNIQUE
);
INSERT INTO event_types (event_type_nid, event_type) VALUES
(1, 'm.room.create'),
(2, 'm.room.power_levels'),
(3, 'm.room.join_rules'),
(4, 'm.room.third_party_invite'),
(5, 'm.room.member'),
(6, 'm.room.redaction'),
(7, 'm.room.history_visibility') ON CONFLICT DO NOTHING;
`
// Assign a new numeric event type ID.
// The usual case is that the event type is not in the database.
// In that case the ID will be assigned using the next value from the sequence.
// We use `RETURNING` to tell postgres to return the assigned ID.
// But it's possible that the type was added in a query that raced with us.
// This will result in a conflict on the event_type_unique constraint.
// We peform a update that does nothing rather that doing nothing at all because
// postgres won't return anything unless we touch a row in the table.
const insertEventTypeNIDSQL = "" +
"INSERT INTO event_types (event_type) VALUES ($1)" +
" ON CONFLICT ON CONSTRAINT event_type_unique" +
" DO UPDATE SET event_type = $1" +
" RETURNING (event_type_nid)"
const selectEventTypeNIDSQL = "" +
"SELECT event_type_nid FROM event_types WHERE event_type = $1"
type eventTypeStatements struct {
insertEventTypeNIDStmt *sql.Stmt
selectEventTypeNIDStmt *sql.Stmt
}
func (s *eventTypeStatements) prepare(db *sql.DB) (err error) {
_, err = db.Exec(eventTypesSchema)
if err != nil {
return
}
if s.insertEventTypeNIDStmt, err = db.Prepare(insertEventTypeNIDSQL); err != nil {
return
}
if s.selectEventTypeNIDStmt, err = db.Prepare(selectEventTypeNIDSQL); err != nil {
return
}
return
}
func (s *eventTypeStatements) insertEventTypeNID(eventType string) (types.EventTypeNID, error) {
var eventTypeNID int64
err := s.insertEventTypeNIDStmt.QueryRow(eventType).Scan(&eventTypeNID)
return types.EventTypeNID(eventTypeNID), err
}
func (s *eventTypeStatements) selectEventTypeNID(eventType string) (types.EventTypeNID, error) {
var eventTypeNID int64
err := s.selectEventTypeNIDStmt.QueryRow(eventType).Scan(&eventTypeNID)
return types.EventTypeNID(eventTypeNID), err
}

178
src/github.com/matrix-org/dendrite/roomserver/storage/events_table.go Normal file
View file

@ -0,0 +1,178 @@
package storage
import (
"database/sql"
"fmt"
"github.com/lib/pq"
"github.com/matrix-org/dendrite/roomserver/types"
)
const eventsSchema = `
-- The events table holds metadata for each event, the actual JSON is stored
-- separately to keep the size of the rows small.
CREATE SEQUENCE IF NOT EXISTS event_nid_seq;
CREATE TABLE IF NOT EXISTS events (
-- Local numeric ID for the event.
event_nid BIGINT PRIMARY KEY DEFAULT nextval('event_nid_seq'),
-- Local numeric ID for the room the event is in.
-- This is never 0.
room_nid BIGINT NOT NULL,
-- Local numeric ID for the type of the event.
-- This is never 0.
event_type_nid BIGINT NOT NULL,
-- Local numeric ID for the state_key of the event
-- This is 0 if the event is not a state event.
event_state_key_nid BIGINT NOT NULL,
-- Local numeric ID for the state at the event.
-- This is 0 if we don't know the state at the event.
-- If the state is not 0 then this event is part of the contiguous
-- part of the event graph
-- Since many different events can have the same state we store the
-- state into a separate state table and refer to it by numeric ID.
state_snapshot_nid bigint NOT NULL DEFAULT 0,
-- The textual event id.
-- Used to lookup the numeric ID when processing requests.
-- Needed for state resolution.
-- An event may only appear in this table once.
event_id TEXT NOT NULL CONSTRAINT event_id_unique UNIQUE,
-- The sha256 reference hash for the event.
-- Needed for setting reference hashes when sending new events.
reference_sha256 BYTEA NOT NULL,
-- A list of numeric IDs for events that can authenticate this event.
auth_event_nids BIGINT[] NOT NULL
);
`
const insertEventSQL = "" +
"INSERT INTO events (room_nid, event_type_nid, event_state_key_nid, event_id, reference_sha256, auth_event_nids)" +
" VALUES ($1, $2, $3, $4, $5, $6)" +
" ON CONFLICT ON CONSTRAINT event_id_unique" +
" DO UPDATE SET event_id = $1" +
" RETURNING event_nid, state_snapshot_nid"
// Bulk lookup of events by string ID.
// Sort by the numeric IDs for event type and state key.
// This means we can use binary search to lookup entries by type and state key.
const bulkSelectStateEventByIDSQL = "" +
"SELECT event_type_nid, event_state_key_nid, event_nid FROM events" +
" WHERE event_id = ANY($1)" +
" ORDER BY event_type_nid, event_state_key_nid ASC"
const bulkSelectStateAtEventByIDSQL = "" +
"SELECT event_type_nid, event_state_key_nid, event_nid, state_snapshot_nid FROM events" +
" WHERE event_id = ANY($1)"
const updateEventStateSQL = "" +
"UPDATE events SET state_snapshot_nid = $2 WHERE event_nid = $1"
type eventStatements struct {
insertEventStmt *sql.Stmt
bulkSelectStateEventByIDStmt *sql.Stmt
bulkSelectStateAtEventByIDStmt *sql.Stmt
updateEventStateStmt *sql.Stmt
}
func (s *eventStatements) prepare(db *sql.DB) (err error) {
_, err = db.Exec(eventsSchema)
if err != nil {
return
}
if s.insertEventStmt, err = db.Prepare(insertEventSQL); err != nil {
return
}
if s.bulkSelectStateEventByIDStmt, err = db.Prepare(bulkSelectStateEventByIDSQL); err != nil {
return
}
if s.bulkSelectStateAtEventByIDStmt, err = db.Prepare(bulkSelectStateAtEventByIDSQL); err != nil {
return
}
if s.updateEventStateStmt, err = db.Prepare(updateEventStateSQL); err != nil {
return
}
return
}
func (s *eventStatements) insertEvent(
roomNID types.RoomNID, eventTypeNID types.EventTypeNID, eventStateKeyNID types.EventStateKeyNID,
eventID string,
referenceSHA256 []byte,
authEventNIDs []types.EventNID,
) (types.EventNID, types.StateSnapshotNID, error) {
nids := make([]int64, len(authEventNIDs))
for i := range authEventNIDs {
nids[i] = int64(authEventNIDs[i])
}
var eventNID int64
var stateNID int64
err := s.insertEventStmt.QueryRow(
int64(roomNID), int64(eventTypeNID), int64(eventStateKeyNID), eventID, referenceSHA256,
pq.Int64Array(nids),
).Scan(&eventNID, &stateNID)
return types.EventNID(eventNID), types.StateSnapshotNID(stateNID), err
}
func (s *eventStatements) bulkSelectStateEventByID(eventIDs []string) ([]types.StateEntry, error) {
rows, err := s.bulkSelectStateEventByIDStmt.Query(pq.StringArray(eventIDs))
if err != nil {
return nil, err
}
defer rows.Close()
// We know that we will only get as many results as event IDs
// because of the unique constraint on event IDs.
// So we can allocate an array of the correct size now.
// We might get fewer results than IDs so we adjust the length of the slice before returning it.
results := make([]types.StateEntry, len(eventIDs))
i := 0
for ; rows.Next(); i++ {
result := &results[i]
if err = rows.Scan(
&result.EventNID,
&result.EventTypeNID,
&result.EventStateKeyNID,
); err != nil {
return nil, err
}
}
if i != len(eventIDs) {
// If there are fewer rows returned than IDs then we were asked to lookup event IDs we don't have.
// We don't know which ones were missing because we don't return the string IDs in the query.
// However it should be possible debug this by replaying queries or entries from the input kafka logs.
// If this turns out to be impossible and we do need the debug information here, it would be better
// to do it as a separate query rather than slowing down/complicating the common case.
return nil, fmt.Errorf("storage: state event IDs missing from the database (%d != %d)", i, len(eventIDs))
}
return results, err
}
func (s *eventStatements) bulkSelectStateAtEventByID(eventIDs []string) ([]types.StateAtEvent, error) {
rows, err := s.bulkSelectStateAtEventByIDStmt.Query(pq.StringArray(eventIDs))
if err != nil {
return nil, err
}
defer rows.Close()
results := make([]types.StateAtEvent, len(eventIDs))
i := 0
for ; rows.Next(); i++ {
result := &results[i]
if err = rows.Scan(
&result.EventNID,
&result.EventTypeNID,
&result.EventStateKeyNID,
&result.BeforeStateSnapshotNID,
); err != nil {
return nil, err
}
if result.BeforeStateSnapshotNID == 0 {
return nil, fmt.Errorf("storage: missing state for event NID %d", result.EventNID)
}
}
if i != len(eventIDs) {
return nil, fmt.Errorf("storage: event IDs missing from the database (%d != %d)", i, len(eventIDs))
}
return results, err
}
func (s *eventStatements) updateEventState(eventNID types.EventNID, stateNID types.StateSnapshotNID) error {
_, err := s.updateEventStateStmt.Exec(int64(eventNID), int64(stateNID))
return err
}

67
src/github.com/matrix-org/dendrite/roomserver/storage/partition_offset_table.go Normal file
View file

@ -0,0 +1,67 @@
package storage
import (
"database/sql"
"github.com/matrix-org/dendrite/roomserver/types"
)
const partitionOffsetsSchema = `
-- The offsets that the server has processed up to.
CREATE TABLE IF NOT EXISTS partition_offsets (
-- The name of the topic.
topic TEXT NOT NULL,
-- The 32-bit partition ID
partition INTEGER NOT NULL,
-- The 64-bit offset.
partition_offset BIGINT NOT NULL,
CONSTRAINT topic_partition_unique UNIQUE (topic, partition)
);
`
const selectPartitionOffsetsSQL = "" +
"SELECT partition, partition_offset FROM partition_offsets WHERE topic = $1"
const upsertPartitionOffsetsSQL = "" +
"INSERT INTO partition_offsets (topic, partition, partition_offset) VALUES ($1, $2, $3)" +
" ON CONFLICT ON CONSTRAINT topic_partition_unique" +
" DO UPDATE SET partition_offset = $3"
type partitionOffsetStatements struct {
selectPartitionOffsetsStmt *sql.Stmt
upsertPartitionOffsetStmt *sql.Stmt
}
func (s *partitionOffsetStatements) prepare(db *sql.DB) (err error) {
_, err = db.Exec(partitionOffsetsSchema)
if err != nil {
return
}
if s.selectPartitionOffsetsStmt, err = db.Prepare(selectPartitionOffsetsSQL); err != nil {
return
}
if s.upsertPartitionOffsetStmt, err = db.Prepare(upsertPartitionOffsetsSQL); err != nil {
return
}
return
}
func (s *partitionOffsetStatements) selectPartitionOffsets(topic string) ([]types.PartitionOffset, error) {
rows, err := s.selectPartitionOffsetsStmt.Query(topic)
if err != nil {
return nil, err
}
defer rows.Close()
var results []types.PartitionOffset
for rows.Next() {
var offset types.PartitionOffset
if err := rows.Scan(&offset.Partition, &offset.Offset); err != nil {
return nil, err
}
}
return results, nil
}
func (s *partitionOffsetStatements) upsertPartitionOffset(topic string, partition int32, offset int64) error {
_, err := s.upsertPartitionOffsetStmt.Exec(topic, partition, offset)
return err
}

57
src/github.com/matrix-org/dendrite/roomserver/storage/rooms_table.go Normal file
View file

@ -0,0 +1,57 @@
package storage
import (
"database/sql"
"github.com/matrix-org/dendrite/roomserver/types"
)
const roomsSchema = `
CREATE SEQUENCE IF NOT EXISTS room_nid_seq;
CREATE TABLE IF NOT EXISTS rooms (
-- Local numeric ID for the room.
room_nid BIGINT PRIMARY KEY DEFAULT nextval('room_nid_seq'),
-- Textual ID for the room.
room_id TEXT NOT NULL CONSTRAINT room_id_unique UNIQUE
);
`
// Same as insertEventTypeNIDSQL
const insertRoomNIDSQL = "" +
"INSERT INTO rooms (room_id) VALUES ($1)" +
" ON CONFLICT ON CONSTRAINT room_id_unique" +
" DO UPDATE SET room_id = $1" +
" RETURNING (room_nid)"
const selectRoomNIDSQL = "" +
"SELECT room_nid FROM rooms WHERE room_id = $1"
type roomStatements struct {
insertRoomNIDStmt *sql.Stmt
selectRoomNIDStmt *sql.Stmt
}
func (s *roomStatements) prepare(db *sql.DB) (err error) {
_, err = db.Exec(roomsSchema)
if err != nil {
return
}
if s.insertRoomNIDStmt, err = db.Prepare(insertRoomNIDSQL); err != nil {
return
}
if s.selectRoomNIDStmt, err = db.Prepare(selectRoomNIDSQL); err != nil {
return
}
return
}
func (s *roomStatements) insertRoomNID(roomID string) (types.RoomNID, error) {
var roomNID int64
err := s.insertRoomNIDStmt.QueryRow(roomID).Scan(&roomNID)
return types.RoomNID(roomNID), err
}
func (s *roomStatements) selectRoomNID(roomID string) (types.RoomNID, error) {
var roomNID int64
err := s.selectRoomNIDStmt.QueryRow(roomID).Scan(&roomNID)
return types.RoomNID(roomNID), err
}

776
src/github.com/matrix-org/dendrite/roomserver/storage/sql.go
View file

@ -2,785 +2,53 @@ package storage
import (
"database/sql"
"fmt"
"github.com/lib/pq"
"github.com/matrix-org/dendrite/roomserver/types"
)
type statements struct {
selectPartitionOffsetsStmt *sql.Stmt
upsertPartitionOffsetStmt *sql.Stmt
insertEventTypeNIDStmt *sql.Stmt
selectEventTypeNIDStmt *sql.Stmt
insertEventStateKeyNIDStmt *sql.Stmt
selectEventStateKeyNIDStmt *sql.Stmt
bulkSelectEventStateKeyNIDStmt *sql.Stmt
insertRoomNIDStmt *sql.Stmt
selectRoomNIDStmt *sql.Stmt
insertEventStmt *sql.Stmt
bulkSelectStateEventByIDStmt *sql.Stmt
bulkSelectStateAtEventByIDStmt *sql.Stmt
updateEventStateStmt *sql.Stmt
insertEventJSONStmt *sql.Stmt
bulkSelectEventJSONStmt *sql.Stmt
insertStateStmt *sql.Stmt
bulkSelectStateBlockNIDsStmt *sql.Stmt
insertStateDataStmt *sql.Stmt
selectNextStateBlockNIDStmt *sql.Stmt
bulkSelectStateDataEntriesStmt *sql.Stmt
partitionOffsetStatements
eventTypeStatements
eventStateKeyStatements
roomStatements
eventStatements
eventJSONStatements
stateSnapshotStatements
stateBlockStatements
}
func (s *statements) prepare(db *sql.DB) error {
var err error
if err = s.preparePartitionOffsets(db); err != nil {
if err = s.partitionOffsetStatements.prepare(db); err != nil {
return err
}
if err = s.prepareEventTypes(db); err != nil {
if err = s.eventTypeStatements.prepare(db); err != nil {
return err
}
if err = s.prepareEventStateKeys(db); err != nil {
if err = s.eventStateKeyStatements.prepare(db); err != nil {
return err
}
if err = s.prepareRooms(db); err != nil {
if err = s.roomStatements.prepare(db); err != nil {
return err
}
if err = s.prepareEvents(db); err != nil {
if err = s.eventStatements.prepare(db); err != nil {
return err
}
if err = s.prepareEventJSON(db); err != nil {
if err = s.eventJSONStatements.prepare(db); err != nil {
return err
}
if err = s.stateSnapshotStatements.prepare(db); err != nil {
return err
}
if err = s.stateBlockStatements.prepare(db); err != nil {
return err
}
return nil
}
func (s *statements) preparePartitionOffsets(db *sql.DB) (err error) {
_, err = db.Exec(partitionOffsetsSchema)
if err != nil {
return
}
if s.selectPartitionOffsetsStmt, err = db.Prepare(selectPartitionOffsetsSQL); err != nil {
return
}
if s.upsertPartitionOffsetStmt, err = db.Prepare(upsertPartitionOffsetsSQL); err != nil {
return
}
return
}
const partitionOffsetsSchema = `
-- The offsets that the server has processed up to.
CREATE TABLE IF NOT EXISTS partition_offsets (
-- The name of the topic.
topic TEXT NOT NULL,
-- The 32-bit partition ID
partition INTEGER NOT NULL,
-- The 64-bit offset.
partition_offset BIGINT NOT NULL,
CONSTRAINT topic_partition_unique UNIQUE (topic, partition)
);
`
const selectPartitionOffsetsSQL = "" +
"SELECT partition, partition_offset FROM partition_offsets WHERE topic = $1"
const upsertPartitionOffsetsSQL = "" +
"INSERT INTO partition_offsets (topic, partition, partition_offset) VALUES ($1, $2, $3)" +
" ON CONFLICT ON CONSTRAINT topic_partition_unique" +
" DO UPDATE SET partition_offset = $3"
func (s *statements) selectPartitionOffsets(topic string) ([]types.PartitionOffset, error) {
rows, err := s.selectPartitionOffsetsStmt.Query(topic)
if err != nil {
return nil, err
}
defer rows.Close()
var results []types.PartitionOffset
for rows.Next() {
var offset types.PartitionOffset
if err := rows.Scan(&offset.Partition, &offset.Offset); err != nil {
return nil, err
}
}
return results, nil
}
func (s *statements) upsertPartitionOffset(topic string, partition int32, offset int64) error {
_, err := s.upsertPartitionOffsetStmt.Exec(topic, partition, offset)
return err
}
func (s *statements) prepareEventTypes(db *sql.DB) (err error) {
_, err = db.Exec(eventTypesSchema)
if err != nil {
return
}
if s.insertEventTypeNIDStmt, err = db.Prepare(insertEventTypeNIDSQL); err != nil {
return
}
if s.selectEventTypeNIDStmt, err = db.Prepare(selectEventTypeNIDSQL); err != nil {
return
}
return
}
const eventTypesSchema = `
-- Numeric versions of the event "type"s. Event types tend to be taken from a
-- small common pool. Assigning each a numeric ID should reduce the amount of
-- data that needs to be stored and fetched from the database.
-- It also means that many operations can work with int64 arrays rather than
-- string arrays which may help reduce GC pressure.
-- Well known event types are pre-assigned numeric IDs:
-- 1 -> m.room.create
-- 2 -> m.room.power_levels
-- 3 -> m.room.join_rules
-- 4 -> m.room.third_party_invite
-- 5 -> m.room.member
-- 6 -> m.room.redaction
-- 7 -> m.room.history_visibility
-- Picking well-known numeric IDs for the events types that require special
-- attention during state conflict resolution means that we write that code
-- using numeric constants.
-- It also means that the numeric IDs for common event types should be
-- consistent between different instances which might make ad-hoc debugging
-- easier.
-- Other event types are automatically assigned numeric IDs starting from 2**16.
-- This leaves room to add more pre-assigned numeric IDs and clearly separates
-- the automatically assigned IDs from the pre-assigned IDs.
CREATE SEQUENCE IF NOT EXISTS event_type_nid_seq START 65536;
CREATE TABLE IF NOT EXISTS event_types (
-- Local numeric ID for the event type.
event_type_nid BIGINT PRIMARY KEY DEFAULT nextval('event_type_nid_seq'),
-- The string event_type.
event_type TEXT NOT NULL CONSTRAINT event_type_unique UNIQUE
);
INSERT INTO event_types (event_type_nid, event_type) VALUES
(1, 'm.room.create'),
(2, 'm.room.power_levels'),
(3, 'm.room.join_rules'),
(4, 'm.room.third_party_invite'),
(5, 'm.room.member'),
(6, 'm.room.redaction'),
(7, 'm.room.history_visibility') ON CONFLICT DO NOTHING;
`
// Assign a new numeric event type ID.
// The usual case is that the event type is not in the database.
// In that case the ID will be assigned using the next value from the sequence.
// We use `RETURNING` to tell postgres to return the assigned ID.
// But it's possible that the type was added in a query that raced with us.
// This will result in a conflict on the event_type_unique constraint.
// We peform a update that does nothing rather that doing nothing at all because
// postgres won't return anything unless we touch a row in the table.
const insertEventTypeNIDSQL = "" +
"INSERT INTO event_types (event_type) VALUES ($1)" +
" ON CONFLICT ON CONSTRAINT event_type_unique" +
" DO UPDATE SET event_type = $1" +
" RETURNING (event_type_nid)"
const selectEventTypeNIDSQL = "" +
"SELECT event_type_nid FROM event_types WHERE event_type = $1"
func (s *statements) insertEventTypeNID(eventType string) (types.EventTypeNID, error) {
var eventTypeNID int64
err := s.insertEventTypeNIDStmt.QueryRow(eventType).Scan(&eventTypeNID)
return types.EventTypeNID(eventTypeNID), err
}
func (s *statements) selectEventTypeNID(eventType string) (types.EventTypeNID, error) {
var eventTypeNID int64
err := s.selectEventTypeNIDStmt.QueryRow(eventType).Scan(&eventTypeNID)
return types.EventTypeNID(eventTypeNID), err
}
func (s *statements) prepareEventStateKeys(db *sql.DB) (err error) {
_, err = db.Exec(eventStateKeysSchema)
if err != nil {
return
}
if s.insertEventStateKeyNIDStmt, err = db.Prepare(insertEventStateKeyNIDSQL); err != nil {
return
}
if s.selectEventStateKeyNIDStmt, err = db.Prepare(selectEventStateKeyNIDSQL); err != nil {
return
}
if s.bulkSelectEventStateKeyNIDStmt, err = db.Prepare(bulkSelectEventStateKeyNIDSQL); err != nil {
return
}
return
}
const eventStateKeysSchema = `
-- Numeric versions of the event "state_key"s. State keys tend to be reused so
-- assigning each string a numeric ID should reduce the amount of data that
-- needs to be stored and fetched from the database.
-- It also means that many operations can work with int64 arrays rather than
-- string arrays which may help reduce GC pressure.
-- Well known state keys are pre-assigned numeric IDs:
-- 1 -> "" (the empty string)
-- Other state keys are automatically assigned numeric IDs starting from 2**16.
-- This leaves room to add more pre-assigned numeric IDs and clearly separates
-- the automatically assigned IDs from the pre-assigned IDs.
CREATE SEQUENCE IF NOT EXISTS event_state_key_nid_seq START 65536;
CREATE TABLE IF NOT EXISTS event_state_keys (
-- Local numeric ID for the state key.
event_state_key_nid BIGINT PRIMARY KEY DEFAULT nextval('event_state_key_nid_seq'),
event_state_key TEXT NOT NULL CONSTRAINT event_state_key_unique UNIQUE
);
INSERT INTO event_state_keys (event_state_key_nid, event_state_key) VALUES
(1, '') ON CONFLICT DO NOTHING;
`
// Same as insertEventTypeNIDSQL
const insertEventStateKeyNIDSQL = "" +
"INSERT INTO event_state_keys (event_state_key) VALUES ($1)" +
" ON CONFLICT ON CONSTRAINT event_state_key_unique" +
" DO UPDATE SET event_state_key = $1" +
" RETURNING (event_state_key_nid)"
const selectEventStateKeyNIDSQL = "" +
"SELECT event_state_key_nid FROM event_state_keys WHERE event_state_key = $1"
// Bulk lookup from string state key to numeric ID for that state key.
// Takes an array of strings as the query parameter.
const bulkSelectEventStateKeyNIDSQL = "" +
"SELECT event_state_key, event_state_key_nid FROM event_state_keys" +
" WHERE event_state_key = ANY($1)"
func (s *statements) insertEventStateKeyNID(eventStateKey string) (types.EventStateKeyNID, error) {
var eventStateKeyNID int64
err := s.insertEventStateKeyNIDStmt.QueryRow(eventStateKey).Scan(&eventStateKeyNID)
return types.EventStateKeyNID(eventStateKeyNID), err
}
func (s *statements) selectEventStateKeyNID(eventStateKey string) (types.EventStateKeyNID, error) {
var eventStateKeyNID int64
err := s.selectEventStateKeyNIDStmt.QueryRow(eventStateKey).Scan(&eventStateKeyNID)
return types.EventStateKeyNID(eventStateKeyNID), err
}
func (s *statements) bulkSelectEventStateKeyNID(eventStateKeys []string) (map[string]types.EventStateKeyNID, error) {
rows, err := s.bulkSelectEventStateKeyNIDStmt.Query(pq.StringArray(eventStateKeys))
if err != nil {
return nil, err
}
defer rows.Close()
result := make(map[string]types.EventStateKeyNID, len(eventStateKeys))
for rows.Next() {
var stateKey string
var stateKeyNID int64
if err := rows.Scan(&stateKey, &stateKeyNID); err != nil {
return nil, err
}
result[stateKey] = types.EventStateKeyNID(stateKeyNID)
}
return result, nil
}
func (s *statements) prepareRooms(db *sql.DB) (err error) {
_, err = db.Exec(roomsSchema)
if err != nil {
return
}
if s.insertRoomNIDStmt, err = db.Prepare(insertRoomNIDSQL); err != nil {
return
}
if s.selectRoomNIDStmt, err = db.Prepare(selectRoomNIDSQL); err != nil {
return
}
return
}
const roomsSchema = `
CREATE SEQUENCE IF NOT EXISTS room_nid_seq;
CREATE TABLE IF NOT EXISTS rooms (
-- Local numeric ID for the room.
room_nid BIGINT PRIMARY KEY DEFAULT nextval('room_nid_seq'),
-- Textual ID for the room.
room_id TEXT NOT NULL CONSTRAINT room_id_unique UNIQUE
);
`
// Same as insertEventTypeNIDSQL
const insertRoomNIDSQL = "" +
"INSERT INTO rooms (room_id) VALUES ($1)" +
" ON CONFLICT ON CONSTRAINT room_id_unique" +
" DO UPDATE SET room_id = $1" +
" RETURNING (room_nid)"
const selectRoomNIDSQL = "" +
"SELECT room_nid FROM rooms WHERE room_id = $1"
func (s *statements) insertRoomNID(roomID string) (types.RoomNID, error) {
var roomNID int64
err := s.insertRoomNIDStmt.QueryRow(roomID).Scan(&roomNID)
return types.RoomNID(roomNID), err
}
func (s *statements) selectRoomNID(roomID string) (types.RoomNID, error) {
var roomNID int64
err := s.selectRoomNIDStmt.QueryRow(roomID).Scan(&roomNID)
return types.RoomNID(roomNID), err
}
const eventsSchema = `
-- The events table holds metadata for each event, the actual JSON is stored
-- separately to keep the size of the rows small.
CREATE SEQUENCE IF NOT EXISTS event_nid_seq;
CREATE TABLE IF NOT EXISTS events (
-- Local numeric ID for the event.
event_nid BIGINT PRIMARY KEY DEFAULT nextval('event_nid_seq'),
-- Local numeric ID for the room the event is in.
-- This is never 0.
room_nid BIGINT NOT NULL,
-- Local numeric ID for the type of the event.
-- This is never 0.
event_type_nid BIGINT NOT NULL,
-- Local numeric ID for the state_key of the event
-- This is 0 if the event is not a state event.
event_state_key_nid BIGINT NOT NULL,
-- Local numeric ID for the state at the event.
-- This is 0 if we don't know the state at the event.
-- If the state is not 0 then this event is part of the contiguous
-- part of the event graph
-- Since many different events can have the same state we store the
-- state into a separate state table and refer to it by numeric ID.
state_snapshot_nid bigint NOT NULL DEFAULT 0,
-- The textual event id.
-- Used to lookup the numeric ID when processing requests.
-- Needed for state resolution.
-- An event may only appear in this table once.
event_id TEXT NOT NULL CONSTRAINT event_id_unique UNIQUE,
-- The sha256 reference hash for the event.
-- Needed for setting reference hashes when sending new events.
reference_sha256 BYTEA NOT NULL,
-- A list of numeric IDs for events that can authenticate this event.
auth_event_nids BIGINT[] NOT NULL
);
`
const insertEventSQL = "" +
"INSERT INTO events (room_nid, event_type_nid, event_state_key_nid, event_id, reference_sha256, auth_event_nids)" +
" VALUES ($1, $2, $3, $4, $5, $6)" +
" ON CONFLICT ON CONSTRAINT event_id_unique" +
" DO UPDATE SET event_id = $1" +
" RETURNING event_nid, state_snapshot_nid"
// Bulk lookup of events by string ID.
// Sort by the numeric IDs for event type and state key.
// This means we can use binary search to lookup entries by type and state key.
const bulkSelectStateEventByIDSQL = "" +
"SELECT event_type_nid, event_state_key_nid, event_nid FROM events" +
" WHERE event_id = ANY($1)" +
" ORDER BY event_type_nid, event_state_key_nid ASC"
const bulkSelectStateAtEventByIDSQL = "" +
"SELECT event_type_nid, event_state_key_nid, event_nid, state_snapshot_nid FROM events" +
" WHERE event_id = ANY($1)"
const updateEventStateSQL = "" +
"UPDATE events SET state_snapshot_nid = $2 WHERE event_nid = $1"
func (s *statements) prepareEvents(db *sql.DB) (err error) {
_, err = db.Exec(eventsSchema)
if err != nil {
return
}
if s.insertEventStmt, err = db.Prepare(insertEventSQL); err != nil {
return
}
if s.bulkSelectStateEventByIDStmt, err = db.Prepare(bulkSelectStateEventByIDSQL); err != nil {
return
}
if s.bulkSelectStateAtEventByIDStmt, err = db.Prepare(bulkSelectStateAtEventByIDSQL); err != nil {
return
}
if s.updateEventStateStmt, err = db.Prepare(updateEventStateSQL); err != nil {
return
}
return
}
func (s *statements) insertEvent(
roomNID types.RoomNID, eventTypeNID types.EventTypeNID, eventStateKeyNID types.EventStateKeyNID,
eventID string,
referenceSHA256 []byte,
authEventNIDs []types.EventNID,
) (types.EventNID, types.StateSnapshotNID, error) {
nids := make([]int64, len(authEventNIDs))
for i := range authEventNIDs {
nids[i] = int64(authEventNIDs[i])
}
var eventNID int64
var stateNID int64
err := s.insertEventStmt.QueryRow(
int64(roomNID), int64(eventTypeNID), int64(eventStateKeyNID), eventID, referenceSHA256,
pq.Int64Array(nids),
).Scan(&eventNID, &stateNID)
return types.EventNID(eventNID), types.StateSnapshotNID(stateNID), err
}
func (s *statements) bulkSelectStateEventByID(eventIDs []string) ([]types.StateEntry, error) {
rows, err := s.bulkSelectStateEventByIDStmt.Query(pq.StringArray(eventIDs))
if err != nil {
return nil, err
}
defer rows.Close()
// We know that we will only get as many results as event IDs
// because of the unique constraint on event IDs.
// So we can allocate an array of the correct size now.
// We might get fewer results than IDs so we adjust the length of the slice before returning it.
results := make([]types.StateEntry, len(eventIDs))
i := 0
for ; rows.Next(); i++ {
result := &results[i]
if err = rows.Scan(
&result.EventNID,
&result.EventTypeNID,
&result.EventStateKeyNID,
); err != nil {
return nil, err
}
}
if i != len(eventIDs) {
// If there are fewer rows returned than IDs then we were asked to lookup event IDs we don't have.
// We don't know which ones were missing because we don't return the string IDs in the query.
// However it should be possible debug this by replaying queries or entries from the input kafka logs.
// If this turns out to be impossible and we do need the debug information here, it would be better
// to do it as a separate query rather than slowing down/complicating the common case.
return nil, fmt.Errorf("storage: state event IDs missing from the database (%d != %d)", i, len(eventIDs))
}
return results, err
}
func (s *statements) bulkSelectStateAtEventByID(eventIDs []string) ([]types.StateAtEvent, error) {
rows, err := s.bulkSelectStateAtEventByIDStmt.Query(pq.StringArray(eventIDs))
if err != nil {
return nil, err
}
defer rows.Close()
results := make([]types.StateAtEvent, len(eventIDs))
i := 0
for ; rows.Next(); i++ {
result := &results[i]
if err = rows.Scan(
&result.EventNID,
&result.EventTypeNID,
&result.EventStateKeyNID,
&result.BeforeStateSnapshotNID,
); err != nil {
return nil, err
}
if result.BeforeStateSnapshotNID == 0 {
return nil, fmt.Errorf("storage: missing state for event NID %d", result.EventNID)
}
}
if i != len(eventIDs) {
return nil, fmt.Errorf("storage: event IDs missing from the database (%d != %d)", i, len(eventIDs))
}
return results, err
}
func (s *statements) updateEventState(eventNID types.EventNID, stateNID types.StateSnapshotNID) error {
_, err := s.updateEventStateStmt.Exec(int64(eventNID), int64(stateNID))
return err
}
func (s *statements) prepareEventJSON(db *sql.DB) (err error) {
_, err = db.Exec(eventJSONSchema)
if err != nil {
return
}
if s.insertEventJSONStmt, err = db.Prepare(insertEventJSONSQL); err != nil {
return
}
if s.bulkSelectEventJSONStmt, err = db.Prepare(bulkSelectEventJSONSQL); err != nil {
return
}
return
}
const eventJSONSchema = `
-- Stores the JSON for each event. This kept separate from the main events
-- table to keep the rows in the main events table small.
CREATE TABLE IF NOT EXISTS event_json (
-- Local numeric ID for the event.
event_nid BIGINT NOT NULL PRIMARY KEY,
-- The JSON for the event.
-- Stored as TEXT because this should be valid UTF-8.
-- Not stored as a JSONB because we always just pull the entire event
-- so there is no point in postgres parsing it.
-- Not stored as JSON because we already validate the JSON in the server
-- so there is no point in postgres validating it.
-- TODO: Should we be compressing the events with Snappy or DEFLATE?
event_json TEXT NOT NULL
);
`
const insertEventJSONSQL = "" +
"INSERT INTO event_json (event_nid, event_json) VALUES ($1, $2)" +
" ON CONFLICT DO NOTHING"
// Bulk event JSON lookup by numeric event ID.
// Sort by the numeric event ID.
// This means that we can use binary search to lookup by numeric event ID.
const bulkSelectEventJSONSQL = "" +
"SELECT event_nid, event_json FROM event_json" +
" WHERE event_nid = ANY($1)" +
" ORDER BY event_nid ASC"
func (s *statements) insertEventJSON(eventNID types.EventNID, eventJSON []byte) error {
_, err := s.insertEventJSONStmt.Exec(int64(eventNID), eventJSON)
return err
}
type eventJSONPair struct {
EventNID types.EventNID
EventJSON []byte
}
func (s *statements) bulkSelectEventJSON(eventNIDs []types.EventNID) ([]eventJSONPair, error) {
nids := make([]int64, len(eventNIDs))
for i := range eventNIDs {
nids[i] = int64(eventNIDs[i])
}
rows, err := s.bulkSelectEventJSONStmt.Query(pq.Int64Array(nids))
if err != nil {
return nil, err
}
defer rows.Close()
// We know that we will only get as many results as event NIDs
// because of the unique constraint on event NIDs.
// So we can allocate an array of the correct size now.
// We might get fewer results than NIDs so we adjust the length of the slice before returning it.
results := make([]eventJSONPair, len(eventNIDs))
i := 0
for ; rows.Next(); i++ {
result := &results[i]
var eventNID int64
if err := rows.Scan(&eventNID, &result.EventJSON); err != nil {
return nil, err
}
result.EventNID = types.EventNID(eventNID)
}
return results[:i], nil
}
const stateSchema = `
-- The state of a room before an event.
-- Stored as a list of state_block entries stored in a separate table.
-- The actual state is constructed by combining all the state_block entries
-- referenced by state_block_nids together. If the same state key tuple appears
-- multiple times then the entry from the later state_block clobbers the earlier
-- entries.
-- This encoding format allows us to implement a delta encoding which is useful
-- because room state tends to accumulate small changes over time. Although if
-- the list of deltas becomes too long it becomes more efficient to encode
-- the full state under single state_block_nid.
CREATE SEQUENCE IF NOT EXISTS state_snapshot_nid_seq;
CREATE TABLE IF NOT EXISTS state_snapshots (
-- Local numeric ID for the state.
state_snapshot_nid bigint PRIMARY KEY DEFAULT nextval('state_snapshot_nid_seq'),
-- Local numeric ID of the room this state is for.
-- Unused in normal operation, but useful for background work or ad-hoc debugging.
room_nid bigint NOT NULL,
-- List of state_block_nids, stored sorted by state_block_nid.
state_block_nids bigint[] NOT NULL
);
`
const insertStateSQL = "" +
"INSERT INTO state_snapshots (room_nid, state_block_nids)" +
" VALUES ($1, $2)" +
" RETURNING state_snapshot_nid"
// Bulk state data NID lookup.
// Sorting by state_snapshot_nid means we can use binary search over the result
// to lookup the state data NIDs for a state snapshot NID.
const bulkSelectStateBlockNIDsSQL = "" +
"SELECT state_snapshot_nid, state_block_nids FROM state_snapshots" +
" WHERE state_snapshot_nid = ANY($1) ORDER BY state_snapshot_nid ASC"
func (s *statements) prepareState(db *sql.DB) (err error) {
_, err = db.Exec(stateSchema)
if err != nil {
return
}
if s.insertStateStmt, err = db.Prepare(insertStateSQL); err != nil {
return
}
if s.bulkSelectStateBlockNIDsStmt, err = db.Prepare(bulkSelectStateBlockNIDsSQL); err != nil {
return
}
return
}
func (s *statements) insertState(roomNID types.RoomNID, stateBlockNIDs []types.StateBlockNID) (stateNID types.StateSnapshotNID, err error) {
nids := make([]int64, len(stateBlockNIDs))
for i := range stateBlockNIDs {
nids[i] = int64(stateBlockNIDs[i])
}
err = s.insertStateStmt.QueryRow(int64(roomNID), pq.Int64Array(nids)).Scan(&stateNID)
return
}
func (s *statements) bulkSelectStateBlockNIDs(stateNIDs []types.StateSnapshotNID) ([]types.StateBlockNIDList, error) {
nids := make([]int64, len(stateNIDs))
for i := range stateNIDs {
nids[i] = int64(stateNIDs[i])
}
rows, err := s.bulkSelectStateBlockNIDsStmt.Query(pq.Int64Array(nids))
if err != nil {
return nil, err
}
defer rows.Close()
results := make([]types.StateBlockNIDList, len(stateNIDs))
i := 0
for ; rows.Next(); i++ {
result := &results[i]
var stateBlockNIDs pq.Int64Array
if err := rows.Scan(&result.StateSnapshotNID, &stateBlockNIDs); err != nil {
return nil, err
}
result.StateBlockNIDs = make([]types.StateBlockNID, len(stateBlockNIDs))
for k := range stateBlockNIDs {
result.StateBlockNIDs[k] = types.StateBlockNID(stateBlockNIDs[k])
}
}
if i != len(stateNIDs) {
return nil, fmt.Errorf("storage: state NIDs missing from the database (%d != %d)", i, len(stateNIDs))
}
return results, nil
}
const stateDataSchema = `
-- The state data map.
-- Designed to give enough information to run the state resolution algorithm
-- without hitting the database in the common case.
-- TODO: Is it worth replacing the unique btree index with a covering index so
-- that postgres could lookup the state using an index-only scan?
-- The type and state_key are included in the index to make it easier to
-- lookup a specific (type, state_key) pair for an event. It also makes it easy
-- to read the state for a given state_block_nid ordered by (type, state_key)
-- which in turn makes it easier to merge state data blocks.
CREATE SEQUENCE IF NOT EXISTS state_block_nid_seq;
CREATE TABLE IF NOT EXISTS state_block (
-- Local numeric ID for this state data.
state_block_nid bigint NOT NULL,
event_type_nid bigint NOT NULL,
event_state_key_nid bigint NOT NULL,
event_nid bigint NOT NULL,
UNIQUE (state_block_nid, event_type_nid, event_state_key_nid)
);
`
const insertStateDataSQL = "" +
"INSERT INTO state_block (state_block_nid, event_type_nid, event_state_key_nid, event_nid)" +
" VALUES ($1, $2, $3, $4)"
const selectNextStateBlockNIDSQL = "" +
"SELECT nextval('state_block_nid_seq')"
// Bulk state lookup by numeric event ID.
// Sort by the state_block_nid, event_type_nid, event_state_key_nid
// This means that all the entries for a given state_block_nid will appear
// together in the list and those entries will sorted by event_type_nid
// and event_state_key_nid. This property makes it easier to merge two
// state data blocks together.
const bulkSelectStateDataEntriesSQL = "" +
"SELECT state_block_nid, event_type_nid, event_state_key_nid, event_nid" +
" FROM state_block WHERE state_block_nid = ANY($1)" +
" ORDER BY state_block_nid, event_type_nid, event_state_key_nid"
func (s *statements) prepareStateData(db *sql.DB) (err error) {
_, err = db.Exec(stateDataSchema)
if err != nil {
return
}
if s.insertStateDataStmt, err = db.Prepare(insertStateDataSQL); err != nil {
return
}
if s.selectNextStateBlockNIDStmt, err = db.Prepare(selectNextStateBlockNIDSQL); err != nil {
return
}
if s.bulkSelectStateDataEntriesStmt, err = db.Prepare(bulkSelectStateDataEntriesSQL); err != nil {
return
}
return
}
func (s *statements) bulkInsertStateData(stateBlockNID types.StateBlockNID, entries []types.StateEntry) error {
for _, entry := range entries {
_, err := s.insertStateDataStmt.Exec(
int64(stateBlockNID),
int64(entry.EventTypeNID),
int64(entry.EventStateKeyNID),
int64(entry.EventNID),
)
if err != nil {
return err
}
}
return nil
}
func (s *statements) selectNextStateBlockNID() (types.StateBlockNID, error) {
var stateBlockNID int64
err := s.selectNextStateBlockNIDStmt.QueryRow().Scan(&stateBlockNID)
return types.StateBlockNID(stateBlockNID), err
}
func (s *statements) bulkSelectStateDataEntries(stateBlockNIDs []types.StateBlockNID) ([]types.StateEntryList, error) {
nids := make([]int64, len(stateBlockNIDs))
for i := range stateBlockNIDs {
nids[i] = int64(stateBlockNIDs[i])
}
rows, err := s.bulkSelectStateDataEntriesStmt.Query(pq.Int64Array(nids))
if err != nil {
return nil, err
}
defer rows.Close()
results := make([]types.StateEntryList, len(stateBlockNIDs))
// current is a pointer to the StateEntryList to append the state entries to.
var current *types.StateEntryList
i := 0
for rows.Next() {
var (
stateBlockNID int64
eventTypeNID int64
eventStateKeyNID int64
eventNID int64
entry types.StateEntry
)
if err := rows.Scan(
&stateBlockNID, &eventTypeNID, &eventStateKeyNID, &eventNID,
); err != nil {
return nil, err
}
entry.EventTypeNID = types.EventTypeNID(eventTypeNID)
entry.EventStateKeyNID = types.EventStateKeyNID(eventStateKeyNID)
entry.EventNID = types.EventNID(eventNID)
if current == nil || types.StateBlockNID(stateBlockNID) != current.StateBlockNID {
// The state entry row is for a different state data block to the current one.
// So we start appending to the next entry in the list.
current = &results[i]
current.StateBlockNID = types.StateBlockNID(stateBlockNID)
i++
}
current.StateEntries = append(current.StateEntries, entry)
}
if i != len(stateBlockNIDs) {
return nil, fmt.Errorf("storage: state data NIDs missing from the database (%d != %d)", i, len(stateBlockNIDs))
}
return results, nil
}

137
src/github.com/matrix-org/dendrite/roomserver/storage/state_block_table.go Normal file
View file

@ -0,0 +1,137 @@
package storage
import (
"database/sql"
"fmt"
"github.com/lib/pq"
"github.com/matrix-org/dendrite/roomserver/types"
)
const stateDataSchema = `
-- The state data map.
-- Designed to give enough information to run the state resolution algorithm
-- without hitting the database in the common case.
-- TODO: Is it worth replacing the unique btree index with a covering index so
-- that postgres could lookup the state using an index-only scan?
-- The type and state_key are included in the index to make it easier to
-- lookup a specific (type, state_key) pair for an event. It also makes it easy
-- to read the state for a given state_block_nid ordered by (type, state_key)
-- which in turn makes it easier to merge state data blocks.
CREATE SEQUENCE IF NOT EXISTS state_block_nid_seq;
CREATE TABLE IF NOT EXISTS state_block (
-- Local numeric ID for this state data.
state_block_nid bigint NOT NULL,
event_type_nid bigint NOT NULL,
event_state_key_nid bigint NOT NULL,
event_nid bigint NOT NULL,
UNIQUE (state_block_nid, event_type_nid, event_state_key_nid)
);
`
const insertStateDataSQL = "" +
"INSERT INTO state_block (state_block_nid, event_type_nid, event_state_key_nid, event_nid)" +
" VALUES ($1, $2, $3, $4)"
const selectNextStateBlockNIDSQL = "" +
"SELECT nextval('state_block_nid_seq')"
// Bulk state lookup by numeric event ID.
// Sort by the state_block_nid, event_type_nid, event_state_key_nid
// This means that all the entries for a given state_block_nid will appear
// together in the list and those entries will sorted by event_type_nid
// and event_state_key_nid. This property makes it easier to merge two
// state data blocks together.
const bulkSelectStateDataEntriesSQL = "" +
"SELECT state_block_nid, event_type_nid, event_state_key_nid, event_nid" +
" FROM state_block WHERE state_block_nid = ANY($1)" +
" ORDER BY state_block_nid, event_type_nid, event_state_key_nid"
type stateBlockStatements struct {
insertStateDataStmt *sql.Stmt
selectNextStateBlockNIDStmt *sql.Stmt
bulkSelectStateDataEntriesStmt *sql.Stmt
}
func (s *stateBlockStatements) prepare(db *sql.DB) (err error) {
_, err = db.Exec(stateDataSchema)
if err != nil {
return
}
if s.insertStateDataStmt, err = db.Prepare(insertStateDataSQL); err != nil {
return
}
if s.selectNextStateBlockNIDStmt, err = db.Prepare(selectNextStateBlockNIDSQL); err != nil {
return
}
if s.bulkSelectStateDataEntriesStmt, err = db.Prepare(bulkSelectStateDataEntriesSQL); err != nil {
return
}
return
}
func (s *stateBlockStatements) bulkInsertStateData(stateBlockNID types.StateBlockNID, entries []types.StateEntry) error {
for _, entry := range entries {
_, err := s.insertStateDataStmt.Exec(
int64(stateBlockNID),
int64(entry.EventTypeNID),
int64(entry.EventStateKeyNID),
int64(entry.EventNID),
)
if err != nil {
return err
}
}
return nil
}
func (s *stateBlockStatements) selectNextStateBlockNID() (types.StateBlockNID, error) {
var stateBlockNID int64
err := s.selectNextStateBlockNIDStmt.QueryRow().Scan(&stateBlockNID)
return types.StateBlockNID(stateBlockNID), err
}
func (s *stateBlockStatements) bulkSelectStateDataEntries(stateBlockNIDs []types.StateBlockNID) ([]types.StateEntryList, error) {
nids := make([]int64, len(stateBlockNIDs))
for i := range stateBlockNIDs {
nids[i] = int64(stateBlockNIDs[i])
}
rows, err := s.bulkSelectStateDataEntriesStmt.Query(pq.Int64Array(nids))
if err != nil {
return nil, err
}
defer rows.Close()
results := make([]types.StateEntryList, len(stateBlockNIDs))
// current is a pointer to the StateEntryList to append the state entries to.
var current *types.StateEntryList
i := 0
for rows.Next() {
var (
stateBlockNID int64
eventTypeNID int64
eventStateKeyNID int64
eventNID int64
entry types.StateEntry
)
if err := rows.Scan(
&stateBlockNID, &eventTypeNID, &eventStateKeyNID, &eventNID,
); err != nil {
return nil, err
}
entry.EventTypeNID = types.EventTypeNID(eventTypeNID)
entry.EventStateKeyNID = types.EventStateKeyNID(eventStateKeyNID)
entry.EventNID = types.EventNID(eventNID)
if current == nil || types.StateBlockNID(stateBlockNID) != current.StateBlockNID {
// The state entry row is for a different state data block to the current one.
// So we start appending to the next entry in the list.
current = &results[i]
current.StateBlockNID = types.StateBlockNID(stateBlockNID)
i++
}
current.StateEntries = append(current.StateEntries, entry)
}
if i != len(stateBlockNIDs) {
return nil, fmt.Errorf("storage: state data NIDs missing from the database (%d != %d)", i, len(stateBlockNIDs))
}
return results, nil
}

100
src/github.com/matrix-org/dendrite/roomserver/storage/state_snapshot_table.go Normal file
View file

@ -0,0 +1,100 @@
package storage
import (
"database/sql"
"fmt"
"github.com/lib/pq"
"github.com/matrix-org/dendrite/roomserver/types"
)
const stateSnapshotSchema = `
-- The state of a room before an event.
-- Stored as a list of state_block entries stored in a separate table.
-- The actual state is constructed by combining all the state_block entries
-- referenced by state_block_nids together. If the same state key tuple appears
-- multiple times then the entry from the later state_block clobbers the earlier
-- entries.
-- This encoding format allows us to implement a delta encoding which is useful
-- because room state tends to accumulate small changes over time. Although if
-- the list of deltas becomes too long it becomes more efficient to encode
-- the full state under single state_block_nid.
CREATE SEQUENCE IF NOT EXISTS state_snapshot_nid_seq;
CREATE TABLE IF NOT EXISTS state_snapshots (
-- Local numeric ID for the state.
state_snapshot_nid bigint PRIMARY KEY DEFAULT nextval('state_snapshot_nid_seq'),
-- Local numeric ID of the room this state is for.
-- Unused in normal operation, but useful for background work or ad-hoc debugging.
room_nid bigint NOT NULL,
-- List of state_block_nids, stored sorted by state_block_nid.
state_block_nids bigint[] NOT NULL
);
`
const insertStateSQL = "" +
"INSERT INTO state_snapshots (room_nid, state_block_nids)" +
" VALUES ($1, $2)" +
" RETURNING state_snapshot_nid"
// Bulk state data NID lookup.
// Sorting by state_snapshot_nid means we can use binary search over the result
// to lookup the state data NIDs for a state snapshot NID.
const bulkSelectStateBlockNIDsSQL = "" +
"SELECT state_snapshot_nid, state_block_nids FROM state_snapshots" +
" WHERE state_snapshot_nid = ANY($1) ORDER BY state_snapshot_nid ASC"
type stateSnapshotStatements struct {
insertStateStmt *sql.Stmt
bulkSelectStateBlockNIDsStmt *sql.Stmt
}
func (s *stateSnapshotStatements) prepare(db *sql.DB) (err error) {
_, err = db.Exec(stateSnapshotSchema)
if err != nil {
return
}
if s.insertStateStmt, err = db.Prepare(insertStateSQL); err != nil {
return
}
if s.bulkSelectStateBlockNIDsStmt, err = db.Prepare(bulkSelectStateBlockNIDsSQL); err != nil {
return
}
return
}
func (s *stateSnapshotStatements) insertState(roomNID types.RoomNID, stateBlockNIDs []types.StateBlockNID) (stateNID types.StateSnapshotNID, err error) {
nids := make([]int64, len(stateBlockNIDs))
for i := range stateBlockNIDs {
nids[i] = int64(stateBlockNIDs[i])
}
err = s.insertStateStmt.QueryRow(int64(roomNID), pq.Int64Array(nids)).Scan(&stateNID)
return
}
func (s *stateSnapshotStatements) bulkSelectStateBlockNIDs(stateNIDs []types.StateSnapshotNID) ([]types.StateBlockNIDList, error) {
nids := make([]int64, len(stateNIDs))
for i := range stateNIDs {
nids[i] = int64(stateNIDs[i])
}
rows, err := s.bulkSelectStateBlockNIDsStmt.Query(pq.Int64Array(nids))
if err != nil {
return nil, err
}
defer rows.Close()
results := make([]types.StateBlockNIDList, len(stateNIDs))
i := 0
for ; rows.Next(); i++ {
result := &results[i]
var stateBlockNIDs pq.Int64Array
if err := rows.Scan(&result.StateSnapshotNID, &stateBlockNIDs); err != nil {
return nil, err
}
result.StateBlockNIDs = make([]types.StateBlockNID, len(stateBlockNIDs))
for k := range stateBlockNIDs {
result.StateBlockNIDs[k] = types.StateBlockNID(stateBlockNIDs[k])
}
}
if i != len(stateNIDs) {
return nil, fmt.Errorf("storage: state NIDs missing from the database (%d != %d)", i, len(stateNIDs))
}
return results, nil
}